Every spring, at some of the most sophisticated accounting firms in the country, the same scene plays out. A senior tax manager finishes calculating a client’s obligations using software refined over four decades. It is fast, precise and expensive. Then, the actual payment process begins and things unravel.
The senior tax manager has to lay out and draft detailed instructions to the client, who may not be overly computer savvy. Then, there are a bunch of follow-up communications to ensure everything went smoothly. This takes about 90 minutes on average, time that could be spent on much higher-value tasks.
Now, that process happens instantly. Here’s how we made it happen:
I’ve been building AI for accountants since before the term needed a marketing department. In 2015 I founded MindBridge, a platform that helped auditors detect anomalies in financial data, the kind of patterns no human reviewer could catch across millions of transactions. That was the first wave of AI in accounting: technology that helped humans see better. The second wave is different. It does the work itself.
Most of what gets written about AI and this profession misses that distinction, usually because it’s written by people who have never sat with a tax manager during the first week of April. The difference is reshaping accounting in three ways that matter.
AI is graduating from analysis to execution
The first generation of accounting AI, the generation I helped build, was analytical. It flagged, scored and surfaced. It could tell an auditor that transaction 4,817 looked nothing like the other 4,816. It could do nothing about it. A brilliant analyst with no hands.
Agentic AI completes workflows end to end, with humans approving rather than executing. And the gap it closes first is the one in my opening scene. The industry automated calculating taxes decades ago. Paying them stayed manual, because every federal, state and local authority runs its own portal with its own login, format and deadline. The most automated profession in business services ran its final mile by hand.
Closing that gap is what my current company, Remitian, does. This past tax season, a full-service accounting firm in Wisconsin ran a good amount of client tax-payment workload through our platform: 194 payments totaling roughly $2.65 million across the IRS and six state tax authorities, with zero late penalties and zero payment errors. The firm estimates it reclaimed about 275 hours of senior tax-manager time. That’s seven work weeks of its most experienced people, returned in a single season.
Before, a partner described logging into each state portal, confirming the payment, keying it in by hand, across multiple states and hundreds of clients, as the single biggest time sink in the process. After, the most common thing I hear is that the firm finally trusts what it sees: every estimate paid, every confirmation in one place and no more wondering whether a payment slipped through.
The talent math is forcing the issue
Technology explains how this shift is happening. Demographics explain why now.
More than 300,000 accountants and auditors have left the field since 2020, a contraction of roughly 17%, according to Bureau of Labor Statistics data. The AICPA says roughly three-quarters of its members will reach retirement age within the next 15 years. And the work hasn’t shrunk by a single form or filing deadline.
Firms aren’t adopting AI because it’s fashionable. They’re adopting it because the alternative is asking their most senior people, the ones clients actually want in the room, to spend tax season on data entry. When a 20-year tax manager’s afternoon disappears into portal logins, the firm is burning its scarcest resource on its least valuable work.
For any executive sizing up AI, in accounting or anywhere else, the better question: what could your best people do with seven weeks back? In a profession where experienced judgment is the binding constraint on growth, that question answers itself.
Ask a managing partner what seven weeks of senior time would buy, and the answer is never more time on the golf course. It is the work they cannot hire their way into fast enough: more advisory relationships, more complex returns handled in house and more room to take on clients without burning out the partners. The hours are not the point. What the hours unlock is. That’s the real value.
Trust infrastructure matters more than the model
In most industries, an AI error is an inconvenience. A chatbot invents a product spec, someone catches it, life goes on. In accounting, an error is a penalty, an audit trigger or a damaged client relationship. The tolerance for “mostly right” is zero, because the entire profession is built on being exactly right.
So the firms succeeding with AI are the ones asking unglamorous questions. How is every action verified? What does the audit trail look like? Who approves what, and where does that approval live? Can we reconstruct exactly what happened, payment by payment, six months from now?
The model is the easy part. The hard engineering lives in everything around it: controls, verification, auditability. A vendor that touches your money or your clients’ compliance should be able to produce a SOC 2 Type 2 report and explain its human-approval architecture in one sentence. If it can’t do both, the demo doesn’t matter.
Accounting spent a century building trust through process. AI doesn’t get to skip the line. The systems that win will earn their way into the workflow the way a new hire does: verifiable before they’re trusted, and trusted before they’re autonomous.
What doesn’t change
For all the transformation ahead, judgment doesn’t automate. Neither does the conversation where a client asks what a number means for their family, their business or their next decade. As the mechanical layer dissolves, what remains is the work accountants trained for in the first place. Nobody studies for the CPA exam to retype numbers into a portal.
The first wave of AI made accountants better analysts. The second frees them to be advisors again.
Solon Angel is co-founder and CEO of Remitian, a Miami-based fintech whose AI-powered platform automates tax payments across federal, state and local jurisdictions.
